Error 400 status code (Gemini)

[Coliinnn]

Hello,

I'm trying to use this proxy with Gemini only (not openrouter). I have set my base URL: BASE_URL=https://generativelanguage.googleapis.com/v1beta/openai/ (according to here) and created and defined my API key. The proxy is running and I've got it configured in Raycast, but trying to use a Gemini Ollama model gives me an error 400 status code (no body).

Raycast is configured with this proxy and can detect the five models I defined.

My models.json:

[
  {
    "name": "Gemini 2.5 Flash",
    "id": "google/gemini-2.5-flash",
    "contextLength": 1000000,
    "capabilities": ["vision", "tools"],
    "temperature": 0
  },
  {
    "name": "Gemini 2.5 Flash Thinking",
    "id": "google/gemini-2.5-flash:thinking",
    "contextLength": 1000000,
    "capabilities": ["vision", "tools"],
    "temperature": 1
  },
  {
    "name": "DeepSeek V3",
    "id": "deepseek/deepseek-chat-v3-0324",
    "contextLength": 128000,
    "capabilities": ["tools"]
  },
  {
    "name": "GPT-4o Mini",
    "id": "openai/gpt-4o-mini",
    "contextLength": 128000,
    "capabilities": ["vision", "tools"]
  },
  {
    "name": "Claude Sonnet 4",
    "id": "anthropic/claude-sonnet-4",
    "contextLength": 200000,
    "capabilities": ["vision", "tools"],
    "temperature": 0.7
  }
]

Docker logs show this:

{"level":30,"time":1752684959349,"pid":19,"hostname":"cb4fdb6b1427","category":"HttpEvent","reqId":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","configWithoutMessages":{"model":"google/gemini-2.5-flash","stream":true,"stream_options":{"include_usage":true},"temperature":0},"msg":"ChatCompletionRequest"}
{"level":30,"time":1752684959546,"pid":19,"hostname":"cb4fdb6b1427","category":"HttpEvent","reqId":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","msg":"ConnectionCleanup"}
{"level":50,"time":1752684959548,"pid":19,"hostname":"cb4fdb6b1427","category":"HttpEvent","reqId":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","err":{"type":"BadRequestError","message":"400 status code (no body)","stack":"Error: 400 status code (no body)\n    at APIError.generate (/app/node_modules/openai/core/error.js:45:20)\n    at OpenAI.makeStatusError (/app/node_modules/openai/client.js:151:32)\n    at OpenAI.makeRequest (/app/node_modules/openai/client.js:293:30)\n    at process.processTicksAndRejections (node:internal/process/task_queues:105:5)\n    at async chatCompletion (/app/dist/controllers/api.js:50:32)","status":400,"headers":{},"requestID":null},"msg":"ErrorHandler"}
{"level":50,"time":1752684959549,"pid":19,"hostname":"cb4fdb6b1427","category":"HttpEvent","reqId":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","req":{"id":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","method":"POST","url":"/api/chat","query":{},"params":{},"headers":{"host":"10.17.7.24:11435","content-type":"application/json","connection":"keep-alive","accept":"application/json","user-agent":"Raycast/0 CFNetwork/3826.500.131 Darwin/24.5.0","content-length":"697","accept-language":"en-GB,en;q=0.9","accept-encoding":"gzip, deflate"},"remoteAddress":"::ffff:10.17.7.11","remotePort":64905},"res":{"statusCode":500,"headers":{"x-powered-by":"Express","x-request-id":"6b58d6f9-ee22-4193-8ae1-5939d19512f3","content-type":"application/json; charset=utf-8","content-length":"79","etag":"W/\"4f-KpVXNRNPplPd6ywl1CWgHirD6rw\""}},"err":{"type":"Error","message":"failed with status code 500","stack":"Error: failed with status code 500\n    at onResFinished (/app/node_modules/pino-http/logger.js:115:39)\n    at ServerResponse.onResponseComplete (/app/node_modules/pino-http/logger.js:178:14)\n    at ServerResponse.emit (node:events:530:35)\n    at onFinish (node:_http_outgoing:1082:10)\n    at callback (node:internal/streams/writable:766:21)\n    at afterWrite (node:internal/streams/writable:710:5)\n    at afterWriteTick (node:internal/streams/writable:696:10)\n    at process.processTicksAndRejections (node:internal/process/task_queues:89:21)"},"responseTime":202,"msg":"request errored"}

But I'm not sure what to make of this. Could you help me out?

Thanks!

[miikkaylisiurunen]

The base URL you have is correct for the Gemini API. The problem is related to the model IDs. The IDs in your models.json are in the format that OpenRouter uses. For the Gemini API, you can find the correct model IDs here. So for 2.5 Flash, it should be gemini-2.5-flash. The model entry should look something like this:

{
  "name": "Gemini 2.5 Flash Thinking (Auto)",
  "id": "gemini-2.5-flash",
  "contextLength": 1000000,
  "capabilities": ["vision", "tools"]
}

The Gemini API has thinking enabled by default for 2.5 Flash, but the thought summaries are not sent by default. If you want to set the thinking effort, you can use the reasoning_effort parameter in the extra field. It supports none, low, medium, or high. Here is an example of how to set it:

{
  "name": "Gemini 2.5 Flash",
  "id": "gemini-2.5-flash",
  "contextLength": 1000000,
  "capabilities": ["vision", "tools"],
  "extra": {
    "reasoning_effort": "none"
  }
}

If you want to see the thought summaries in the response and have more control over the thinking amount, you can use the following configuration:

{
  "name": "Gemini 2.5 Flash Thinking",
  "id": "gemini-2.5-flash",
  "contextLength": 1000000,
  "capabilities": ["vision", "tools"],
  "extra": {
    "extra_body": {
      "google": {
        "thinking_config": {
          "thinking_budget": 20000,
          "include_thoughts": true
        }
      }
    }
  }
}

[Coliinnn]

@miikkaylisiurunen Awesome, that works wonderfully. Thank you!